Synthetic humanitarianism and algorithmic witnessing: generative AI and the politics of humanitarian representation

Humanitarian organisations have begun deploying generative AI imagery in their campaigns, a practice this article terms synthetic humanitarianism. The article asks what conceptual framework is required to analyse the ethical and communicative implications of this practice. Drawing on witnessing theory, decolonial critiques of humanitarian visibility, and technocolonialism, I argue that synthetic humanitarianism intensifies extractive and colonial logics that have structured humanitarian communication since its institutional emergence. I develop the concept of algorithmic witnessing to name the communicative condition in which generative AI systems produce the relations of testimony and moral address without any act of witnessing having taken place and, in the typical case, without the participation of those represented. The framework specifies four dimensions, testimonial fabrication, epistemic extraction, delanguaging, and stereoscopic automation, and I apply it to thirteen documented cases of generative AI image use in humanitarian communications between 2022 and 2025. The analysis demonstrates that these dimensions form a mutually reinforcing system in which the elimination of situated meaning-making enables fabricated testimony and the computational encoding of colonial visual archives produces automated stereotype. I argue that the most significant consequence of synthetic humanitarianism is political rather than epistemological. The displacement of situated meaning-making by algorithmic production tends to eliminate the last structural point at which affected communities could intervene in their own representation.